JBR 2010-07-12
I have a lot of code that writes code. Sometimes the "standard" code that gets written from the config files is not exactly the right thing for some specific case. When this happens I usually modify the standard proc by wrapping a "shim" around it. This little bit of code makes this less crappy. It is less general than the code in stacking but works for me.
proc shim:shift { name } {
if { [info commands ${name}_shimmed] ne {} } { shim:shift ${name}_shimmed }
rename $name ${name}_shimmed
}
proc shim { name args body } {
shim:shift $name
proc $name $args $body
}
proc shim:next { args } { tailcall [lindex [info level -1] 0]_shimmed {*}$args }
Found another version of shim:next even more like stacking. Fixed up to use try instead of catch.
proc shim:next { name args } {
set level -1
try {
while { ![regexp "^${name}(_shimmed)*" [lindex [info level $level] 0]] } { incr level -1 }
} on error message {
error "cannot locate shim: $name"
}
[lindex [info level $level] 0]_shimmed {*}$args
}Using this kind of thing seems to be calling out for replacement with TclOO.